bump to ngff v0.3 (add support for axes)

[joshmoore]

see:

• Data dimensionality and axes metadata ome/ngff#35
• Update reader.py to v0.3 ome/ome-zarr-py#89

bioformats2raw 0.3+ can be upgraded to use the NGFF v0.3 spec. This adds an axes field which permits storing arrays less than 5D if the dimension size <= 1, though bioformats2raw and raw2ometiff can choose to continue store 5D arrays regardless.

cc: @constantinpape

[sbesson]

Discussed briefly today, given the dimension order is stored as 5D in METADATA.ome.xml, I would also assume that having this tool preserving a 5D representation and storing the dimension names under the axes field would be the minimal set of change and also ensure maximum compatibility with downstream utilities like raw2ometiff. Does raw2ometiff makes a hard check on the version of the multiscales? If not, it is possible raw2ometiff 0.3.0 would consume this change out of the box.

This raises the question of whether this issue also encompasses adding full support for the 0.3 specification. Concretely, that would mean a Zarr group with a 0.3.0 bioformats2raw.layout containing individual series using the multiscales 0.3 specification and arrays of less than 5 dimensions. In this case, as suggested by @melissalinkert, it might be better to look at raw2ometiff first.

[sbesson]

Is this issue effectively solved by #113 (jumping directly to the upcoming OME-NGFF 0.4 specification)?

[joshmoore]

I'd think so.

[melissalinkert]

Closing as discussed with @sbesson.